Architectural Design of High-Capacity Cache Based on MRAM Technology

Yifan Yang,Wentao Hong,Zhuang Ma,Lizhou Wu,Yang Ou,Jie Shen,Nong Xiao

DOI: https://doi.org/10.1145/3670105.3670107

2024-01-01

Abstract:In the era of explosive growth in data, the demand for high-capacity cache is ever increasing. Cache system designs need to attain high speed while maintaining acceptable area and energy consumption costs. However, traditional cache systems face challenges such as the slow-down of Moore's Law, large on-chip area occupation of SRAM, and high energy consumption. This paper addresses the need of high-capacity cache systems by proposing a cache architecture based on magnetic random access memory (MRAM). We design and implement an automated cache system simulation framework, conducting automated cache simulation and analysis from the circuit level to the architectural level to evaluate MRAM-based hybrid cache designs. Our architectural-level experiments demonstrate that our proposed cache design, which replaces SRAM with SOT-MRAM for both L1 and L2 caches, shows advantages over traditional SRAM cache designs. It can reduce latency by 29.31%, decrease area occupation by 46.60%, and lower total energy consumption by 23.53%.